Learning to tell two spirals apart

Alexis P. Wieland recently proposed a useful benchmark task for neural networks: distinguishing between two intertwined spirals. Although this task is easy to visualize, it is hard for a network to learn due to its extreme nonlinearity. In this report we exhibit a networkarchitecture that facilitates the learning of the spiral task, and then compare the leaming speed of several variants of the back-propagation algorithm.